神经网络隐含层挖掘大气折射高阶信息的因果论证

大气折射的神经网络建模具有挖掘潜在高阶信息的能力,但其因果关系一直没有很好地显性证明或显性解释.目前的解释是隐性的,不易广为接受.尝试从大气折射映射函数的神经网络变换研究中为前述问题寻找显性论据.首先归纳出映射函数模型的基本形式,为便于分析研究,对它们进行了数学概念上的一般化处理,分析了函数的神经网络变换理论依据,进而建立了映射函数模型对应的神经网络模型.在此基础上,提出了显性证明的方法.最终调用Matlab 7中的神经网络工具箱,在普尔科沃大气折射表数据平台上进行了映射函数与神经网络建模拟合,为论题提供了很好的显性评价依据.

The consequence study of searching the higher order information of atmospheric refraction by the hidden layer units of BP neural network

The reason why the hidden layer units of back propagation neural network(BPNN) has the searching ability of latent higher order information of the atmospheric refraction,has not been proved or explained.This paper attempts to seek a dominant argument for the aforementioned question from the neural network transformation study of the atmospheric refraction mapping function (MF).The fundamental mode of the MF is induced and the generalization formula in mathematics is given.The fundamental transformation theory of the function with neural network is analyzed.The BPNN model for the MF model and the dominant appraisal method are established.The adjustment is carried out with the help of Matlab 7 Neural network toolbox for modeling and fitting.